Staff Machine Learning Engineer, AI Security

Reddit
United States
Workplace: RemoteFull timeUSD 230,000 - 322,000 annuallyFunction: Data Science & Machine LearningExperience: 8+ yearsSkills: ["Communication","Mentorship","Technical leadership","Experiment design","Responsible ML practices"]

Build and deploy ML-powered protections for AI security, focusing on detecting and preventing prompt injection, jailbreaks, sensitive data exposure, and unsafe AI behavior. Select and fine-tune models, create reproducible training/evaluation pipelines, and set the multi-quarter modeling roadmap. Own dataset quality and the production lifecycle using monitoring, red-team feedback, and rigorous evaluation to balance accuracy, latency, reliability, and cost.

Loading

Loading job details...

Preparing the role view and application actions.

FursaFursa
Reddit
Reddit
1 day ago

Staff Machine Learning Engineer, AI Security

✓ Verified Job

Canonical indexed version, validated from employer's careers page.

Source: Company careers pageValidated by: Fursa AI
Last checked: 13 hours agoStatus: Live

Job Summary

Build and deploy ML-powered protections for AI security, focusing on detecting and preventing prompt injection, jailbreaks, sensitive data exposure, and unsafe AI behavior. Select and fine-tune models, create reproducible training/evaluation pipelines, and set the multi-quarter modeling roadmap. Own dataset quality and the production lifecycle using monitoring, red-team feedback, and rigorous evaluation to balance accuracy, latency, reliability, and cost.
Location: United States
Workplace: Remote
Employment Type: Full time
Job Function: Data Science & Machine Learning
Seniority: Mid level

Key Responsibilities

  • •Select, adapt, fine-tune, evaluate, and deploy pretrained models and lightweight classifiers for AI security problems.
  • •Build reproducible training and evaluation pipelines, partnering to improve inference performance, efficiency, and operational reliability.
  • •Set technical vision and a multi-quarter modeling roadmap with cross-functional teams to define architectures and iterate.
  • •Run evaluations and performance analysis to improve accuracy and adversarial robustness, defining launch criteria across false positives, latency, throughput, reliability, and cost.
  • •Own training-data quality and the production model lifecycle using monitoring, incident findings, and red-team feedback to guide dataset improvements and safe rollout or rollback.

Pay and Benefits

Salary: USD 230,000 - 322,000 annually
Equity and Bonus:Equity
Perks:Health InsuranceDentalVision401kPaid LeaveParental Leave

Key Requirements

  • •8+ years building machine learning models with hands-on training experience and measurable production impact.
  • •Strong Python and deep learning framework experience, including TensorFlow, PyTorch, or Hugging Face Transformers.
  • •Expertise in neural network architectures and optimization, including preprocessing, tokenization, embeddings, language modeling, and model calibration.
  • •Experience with scalable/distributed training and data pipelines, such as Ray Train or PyTorch Distributed.
  • •Rigor in experimental design and evaluation (e.g., holdouts, ablations, adversarial tests, error analysis) and strong communication with technical and non-technical partners.
Experience:8+ yearsSecurityTrust and safetyFraudPrivacyAdversarial domains
Skills:CommunicationMentorshipTechnical leadershipExperiment designResponsible ML practices
Languages:English
Tech Stack:PythonTensorFlowPyTorchHugging Face TransformersRay TrainPyTorch DistributedNLPLarge language modelsMultimodal

Company Brief

Reddit
Operates Reddit, a large online community and discussion platform where users submit content, comment, and vote across topic-based communities (subreddits); monetizes via advertising, premium subscriptions, and awards.
Industry: Digital Media
Company Size: Enterprise (1,001+ employees)
Growth: Public Company
Valuation: Public Company (Market Cap in USD)
Funding: IPO / Publicly Listed
Headquarters: San Francisco, United States
Founded: 2005
WebsiteLinkedIn